10/6 



DESIGN FOR A BRAIN 



neurons form not only chains but circuits. Figure 10/5/1 is taken 

 from one of his papers. Such circuits are so common that he has 

 enunciated a c Law of Reciprocity of Connexions ' : ' if a cell- 

 complex A sends fibres to cell or cell-complex B, then B also sends 

 fibres to A, either direct or by means of one internuncial neuron '. 

 A simple circuit, if excited, would tend either to sink back to 

 zero excitation, if the amplification-factor was less than unity, 

 or to rise to maximal excitation if it was greater than unity. 

 Such a circuit tends to maintain only two degrees of activity : 



Figure 10/5/1 : Neurons and their connections in the trigeminal 

 reflex arc. (Semi-diagrammatic ; from Lorente de N6.) 



the inactive and the maximal. Its activity will therefore be of 

 step-function form if the time taken by the chain to build up to 

 maximal excitation can be neglected. Its critical states would 

 be the smallest excitation capable of starting it to full activity, 

 and the smallest inhibition capable of stopping it. McCulloch 

 has referred to such circuits as ' endromes ' and has studied some 

 of their properties. The reader will notice that the ' endrome ' 

 exemplifies the principle of S. 7/4. 



10/6. The definition of the ultrastable system might suggest 

 that an almost infinite number of step-functions is necessary if 

 the system is not to keep repeating itself ; and the reader may 

 wonder whether the nervous system can supply so large a number. 

 In fact the number required is not large. The reason can be 

 shown most simply by a numerical illustration. 



If a step-function can take two values it can provide two fields 

 for the main variables (Figure 7/8/1). If another step-function 

 with two values is added, the total combinations of value are four, 

 and each combination will, in general, produce its own field 

 (S. 21/1). So if there are n step-functions, each capable of taking 



128 



